Modern large language models

A base language model without SFT/RLHF, when fed through a chat template containing a philosophical text priming, produces a coherent, structurally organized, and topically relevant answer to an arbitrary everyday question. The behavior externally resembles an instruction-tuned model, although the model has not undergone such tuning. Protocol 1. Take a base model without instruction training.2. Apply a chat template.3. Add a philosophical text as priming.4. Ask any everyday question.5. Observe: the answer does not degrade into repetition, nonsense, or drift, but stays on topic, structured, and coherent.